Large intelligent surfaces (LISs) is an evolution of massive MIMO systems allowing for huge capacity gains. To reduce implementation complexity, it is convenient to implement the LIS using panels that are either activated or deactivated, and associated to terminals according to their propagation characteristics to the panels. The associated spatial resource allocation to maximise the terminals’, as well as the overall, bit rates can lead to complex optimisation problems.

In this paper we consider resource allocation for panel-based LIS surfaces. We present an iterative sequential algorithm for determining the set of active panels and the respective panel-terminal association for the maximisation of the minimum terminal rate. Our algorithm is decentralised and has low complexity. Moreover, it approaches the performance of much more complex, quasi-optimum algorithms.